text-generation-inference

003163a2 - backend(vllm): map ResultOutput to InferStreamResponse to stream back to the client